The Application of an Improved BPNN Model to Coal Pyrolysis

Abstract

The pyrolysis characteristics of nine kinds of Chinese coals obtained by the thermogravimetry were modeled using artificial neural network. In order to remedy the defects of back propagation algorithm, a momentum term and an adaptive learning rate were introduced. The architecture of the improved back propagation neural network model was 3 × 5 × 3, which included three input nodes (content of volatiles and ash, C/H), five hidden nodes, and three output nodes (the weight loss percentage, the maximum weight loss velocity temperature, and the weight loss peaks). The model prediction results indicated that the modified back propagation neural network had a fast convergence speed and a high accuracy. Moreover, the maximum relative error was less than 4%. Furthermore, in model analysis the effect of each input node was quantized by the sum of absolute values of weights. And the results indicated that C/H had more remarkable influence than the other two.

Improved extreme learning machine based on adaptive dual-strategy optimization algorithm and its application

“…proposed the use of a BP neural network (BPNN) to expose the relationship between the characteristics affecting solar energy and power generation. The neurons in each layer of BPNN are fully connected, and the error is propagated in a reverse step-by-step manner [5][6]. Xiaoqiang et al [7].…”

An Earthquake Fatalities Assessment Method Based on Feature Importance with Deep Learning and Random Forest Models

“…Hence, deep learning method, with its abilities to estimate perplexed relevances, could be an outstanding method to evaluate fatalities. However, BPNN method is not a very perfect network, it has many shortcomings: (1) The convergence speed is too slow and it takes hundreds or more than hundreds of times to learn to converge [22]; (2) it cannot guarantee convergence to a global minimum point [23,24]; (3) there are a number of hidden layers and neurons in that are not theoretically guided, but are determined empirically, thus, the network tends to be large [22]-the redundancy invisibly increases time of network learning [25]; and (4) learning and memory of the network are unstable. Deep learning optimization algorithms can improve the shortages of BPNN method.…”